Reserve Prices Yield Reference Price Effects 75

INTERNATIONAL REAL ESTATE REVIEW

2017 Vol. 20 No. 1: pp. 75 – 104

Do Reserve Prices Yield Reference Price

Effects in Korean Court Auctions of

Residential Real Estate?

Han-Jang No National Pension Service Gyeongin Regional Headquarter, Gyeonggi-do 442-766, South Korea. Phone: 82-10-9132-6410. Email: nhjangnhj@hanmail.net.

Dai-Won Kim School of Urban Planning & Real Estate Studies, Dankook University, Gyeonggi-do 448-701, South Korea. Phone: 82-10-5350-2613. Email: dw2613@gmail.com.

Jung-Suk Yu*

School of Urban Planning & Real Estate Studies, College of Social Science, Dankook University 152, Jukjeon-ro, Suji-gu, Yongin-si, Gyeonggi-do 448-701, South Korea. Phone: 82-31-8005-3338. Email: jsyu@dankook.ac.kr

This study examines whether the reserve prices in court auctions of residential real estate in Seoul, Korea result in reference price effects by influencing the amount of the successful bid. We also explore whether the sensitivity of these reference price effects differ with housing size and assess whether the expected rate of the selling price can be predicted based on the different reserve price levels. The panel data estimates presented herein show that reserve prices positively influence the final property transfer prices; in other words, the reserve prices yield strong reference price effects. The results of the ordinary least square regressions show that the sensitivity of the reference price effects differs with housing size, albeit in an inconsistent manner. Finally, the response surface methodology analysis indicates

* Corresponding author

76 No, Kim and Yu

that different reserve prices lead to different reference price effects with locality across the Seoul metropolitan area. The study thus provides courts and bidders with the means to predict the potential rate of the selling price, which will be useful for decision making in auctions.

Keywords

Reference Price Effect, Reserve Price, Auctions of Residential Real Estate,

Panel Regression, Response Surface Methodology

1. Introduction

Court auctions of residential real estate are ambiguous in nature with

incomplete information so that bidders lack insight in both the value evaluations

made by other bidders and the properties for sale. Despite this information

asymmetry, potential bidders in court auctions of residential real estate in Korea

are compelled to submit their bids quickly because the operation time of an

auction is only an hour or so. Under time constraints, competing bidders are

likely to be psychologically influenced by reserve prices (i.e., minimum prices)

that had been previously announced by the court.

According to the seminal prospect theory put forth by Kahneman and Tversky

(1979), when people respond to stimuli such as brightness, loudness, or

temperature, the past and present contexts of the experience defines the

reference point so that the stimuli are perceived in relation to the reference point.

The seminal prospect theory has convinced behavioral economists to shift the

basis of their insights on auctions from the use of expected utility to reference-

based utility, namely, people are not always rationale and often evaluate utilities

with the use of reference points (Rosenkranz and Schmitz, 2007). For example,

the value perception of bidders in court auctions of residential real estate on the

properties for sale are established with reference to the designated reserve

prices. Accordingly, the reserve prices in all practicality, are used as the

reference point. There is a large volume of theoretical and experimental

research work on these so-called reference price effects, but the results of these

studies are inconclusive in terms of the psychological effect of reserve prices:

whether they are positive, negative, or have no effect (Trautmann and Traxler,

2010).

Against the background provided above, whether the reserve prices of

residential real estate in court administered auctions in Korea result in reference

price effects will be examined in this study. Specifically, we explore whether

reserve prices have a positive psychological influence on final transfer prices,

analyze the sensitivity of the reference price effects of reserve prices to housing

size, and predict the expected selling price based on the different reserve prices.

In order to meet these research objectives, panel data estimates (to assess the

existence of reference price effects), log ordinary least squares (LOLS) (to

Reserve Prices Yield Reference Price Effects 77

determine the sensitivity of the reference price effects), and response surface

methodology (RSM) (to forecast the anticipated selling price) are used in this

study. With the use of these methods, the present paper broadens studies on the

reference price effects of reserve prices by using comparative static analysis

(e.g., static analysis that measures sensitivity and able to make predictions),

whereas the majority of previous work (Ariely and Simonson, 2003; Trautmann

and Traxler, 2010) have only used static analysis to examine reference price

effects. An empirical analysis on first-price sealed-bid auctions on real estate is

novel because most of the literature have only used experimental analysis in

terms of related auctions, or examined real estate auctions in the western

context, or dealt with non-real estate items.

Korea is one of the top three investment venues in the Asia-Pacific region, but

its real estate market is still in the process of maturing.1 This paper will provide

useful information to overseas investors who have growing concerns about the

characteristics of the residential markets of Korea as well as the unique court

auction process. In Korea, court auctions of residential real estate have their

own peculiarities and the process can be divided into three stages: pre-bidding,

actual bidding, and post-bidding. In the pre-bidding stage, the courts must set

the initial reserve prices based on reported values from the appraisers. During

the actual bidding stage, the courts receive sealed bids from the auction

participants with a 10% deposit of the reserve prices. The bidders usually

establish their bids by adding a certain amount of money to the reserve price

based on their own evaluation. The courts then close the bidding process after

the auction has commenced for one hour. The highest bidder wins the auction

and must pay the bidding amount in full within 30 days. If bidders do not meet

the reserve price, the courts continue the process at one-month intervals, and

reduce the reserve price by 20–30% each time. During the court auction process,

bidders do not know the bidding prices of the other competitors. Finally, the

post-bidding stage deals with payment of the purchase price and distribution of

the proceeds after the authorization of the sale (see Appendix and Figure 1).

The residential real estate markets in Korea also have certain unique

characteristics. For example, the dense and crowded environments in Seoul

constitute a characteristic of housing value in that the citizens naturally accept

higher housing price premiums for proximity to subway stations, shopping

malls, and schools. Accordingly, the reverse or nonlinear problem of proximity

premiums for being right next to these spatial entities is not so much an issue

as far as the housing market in Seoul is concerned. The parents and homeowners

in Seoul attach the greatest importance to distance, such as whether their son or

daughter is able to go to school on foot or by bicycle. The prices of high-rise

apartment units on higher floors are more expensive and vice versa. The

traditionally large families and continual increase of house prices have

1 Bae, H.J. “Korean real estate market highly attractive for investors,” The Korean

Herald, July 16, 2013. (http://www.koreaherald.com).

78 No, Kim and Yu

established the tendency to prefer larger dwellings. Yet these may seem strange

to those who are not familiar with the Korean context.

Figure 1 Court Auction Process of Residential Real Estate in Korea

Source: Supreme Court of Korea (http://www.courtauction.go.kr)

The remainder of this paper is organized as follows. Section 2 provides a more

detailed discussion on the reference price effects of reserve prices by way of a

literature review. Section 3 is a description of the methods, data, and variables

used. Section 4 presents the results of an empirical analysis. Section 5

summarizes the findings of this work and provides policy suggestions and

directions for future study.

2. Literature Review

From a theoretical perspective, although early studies on reserve prices did not

fully account for reference price effects, they recognized how reserve prices

influence the property valuation and willingness-to-pay (WTP) of bidders

(Milgrom and Weber, 1982; Vincent, 1995). Mayer (1998) also indicates that

reserve prices are set low in order to provide little information to sellers and not

place binding constraint on bids, even though the units with a published reserve

price should theoretically elicit slightly higher prices. One of the first theoretical

models of reserve prices in auctions is that proposed by Rosenkranz and

Schmitz (2007). They consider the public reserve prices set by auctioneers as

reference points that have significant impacts on bidders and thus have

modified pre-existing expected utility models by assuming that bidders have a

reference-based utility.

Reserve Prices Yield Reference Price Effects 79

The work of Rosenkranz and Schmitz (2007) comprises a stream of research

imported from psychology, termed behavioral economics, which explores the

implications of assuming that human behavior is based on insights. By

supposing that reserve prices have psychological effects on the price-level

decision-making of bidders, behavioral economists assume that price

estimations tend to be made based on publicly announced reserve prices. They

thus recommend that sellers could increase their expected payoff if they set

higher reserve prices because this action subsequently raises the WTP of

bidders, thereby reducing the disparity between the prices paid by the bidders

and the reserve prices set by the sellers.

Similarly, early experimental studies on reserve prices also failed to directly

incorporate reference price effects. Nevertheless, the researchers recognized

their positive influence on successful transfer prices in auctions. For example,

Lin (2005) argues that reserve price has the greatest effect on auction price in

court auctions of residential real estate of the Taipei metropolitan housing

market. Reiley (2005) also indicates that bidders seem to exhibit strategic

behavior to reserve prices, and reserve prices increase the revenues received for

sold goods.

Indeed, there have been increasingly more experimental studies on the

reference price effects of reserve prices. However, the findings of these works

show discrepancies and the overall body of evidence on this topic is

inconclusive. Some authors claim that reserve prices in auctions have positive

psychological influences on the final transfer prices (i.e., starting high and

ending high) (Ariely and Simonson, 2003). Popkowski, Qiu, and He (2009)

focus on the effect of buy-it now prices, which is used as the reference point,

on the WTP of the bidders. Their experiments on the eBay auction site show

that the existence of reserve prices generates a strong reference price effect for

different types of consumer goods, thus resulting in an increase in the WTP of

bidders. Many similar works have also obtained such affirmative results

(Kristensen and Gärling, 1997; Kamins, Dréze, and Folkes, 2004).

However, it is argued in other studies that lower reserve prices ultimately lead

to higher sale prices (i.e., starting low but ending high). For instance, Walley

and Fortin (2005) present a conceptual model of the decision-making process

of consumers in online auctions, and show that a low reserve price results in a

higher final sale price compared with a high reserve price. Ku, Galinsky, and

Murnighan (2006) draw a similar conclusion and argue that this phenomenon

is epitomized by three processes: fewer barriers to entry, escalation of

commitment, and increased value from more traffic. However, a third category

has not managed to draw consistent conclusions about the reference price

effects on reserve prices, that is, studies that focus on the online auction format

(Dewally and Ederington, 2004; Trautmann and Traxler, 2010).

80 No, Kim and Yu

3. Data, Variables, and Methods

3.1 Data Source and Summary Statistics

The time-series and cross-sectional data used in this study consist of 9,631

apartments sold in court auctions of residences in the Seoul metropolitan area

(SMA) from January 2006 to June 2010. We have extracted 9,631 successful

cases out of 25,413 cases posted by five district courts in the city2 because our

study requires information on the sold prices. The average annual number of

apartment auctions in the SMA is 6,062, which comprises 7.1% of the total

number of apartments in the country (i.e., 85,953 apartments).

In the analysis of the behavior of the residential markets, the criterion for

segmenting subordinate markets is influential because the research outcomes

could be affected resultant of the demarcation of this subdivision (Bourassa et

al., 1999). Nonetheless, previous research on court auctions of residential real

estate has subdivided markets in accordance with the jurisdictions of the district

courts, which thus fails to reflect the possible differences in the preference for

a district, the quality, and the price of housing within this district. To overcome

this shortcoming, we have subdivided the study area into five zones by

following the 2020 City Planning of the SMA: central, northeastern,

southeastern, northwestern, and southwestern (Figure 2 and Table 1).

Figure 2 Administrative Districts of the SMA

Source: Seoul Metropolitan Government. (2013).

http://english.seoul.go.kr/gtk/cg/cityhall.php?pidx=5.

Note: There are 25 autonomous districts in the SMA. In the Vision 2020 of Seoul City,

these 25 autonomous districts are classified into five zones.

2 In Seoul, there are five district courts (Central, Northern, Eastern, Western, and

Southern). The jurisdictions of these district courts are divided in accordance with

formal administrative divisions.

Reserve Prices Yield Reference Price Effects 81

Table 1 Five Zones of the SMA

Zone Main Function Autonomous Districts

Central zone International finance

center, central business

center

Jung-gu, Jongno-gu, Yongsan-gu

Northeastern

Zone

Old residential district,

base for employment

Dongdaemun-gu, Seongdong-gu,

Gwangjin-gu, Jungrang-gu, Seongbuk-

gu, Gangbuk-gu, Dobong-gu, Nowon-

gu

Southeastern

Zone

International business,

venture, IT

Seocho-gu, Gangnam-gu, Songpa-gu,

Gangdong-gu

Northwestern

Zone

Nature-friendly

residential district

Eunpyeong-gu, Seodaemun-gu, Mapo-

gu

Southwestern

Zone

High-tech industry,

physical distribution

Gangseo-gu, Yangcheon-gu, Guro-gu,

Geumcheon-gu, Yeongdeungpo-gu,

Dongjak-gu, Gwanak-gu

Source: Seoul Metropolitan Government. (2013).

http://urban.seoul.go.kr/DUPIS/sub3/sub3_1.jsp.

Note: The 2020 City Master plan of the SMA was established under the regulation of

the Act on the Planning and Use of the National Territory. This plan is effective

for 20 years from 2000 to 2020, although it can be amended every five years.

We then classify our data sample into these five categories as well as the SMA

as a whole. Differences in the scope and local characteristics between the five

zones are largely responsible for the considerable discrepancies in the number

of apartments. The central zone, for example, is in the relatively small central

business district of the SMA, while the southwestern and northeastern zones

are the traditional residential areas from the sub-center to the outskirts of the

city. Table 2 presents the summary statistics by zone.

Table 2 Summary Statistics by Zone

Stats SOL RES BID SQU TFL FLO

SMA

Mean 89.53 78.25 6.36 92.23 14.20 7.00

S.D. 11.47 9.71 4.03 25.97 4.24 3.25

Min 22.67 17.00 1.00 13.31 5.00 -1.00

Max 144.25 100.00 44.00 352.02 29.00 22.00

Central Zone

Mean 90.42 78.29 5.50 102.29 13.02 6.91

S.D. 16.48 12.40 4.98 44.26 5.11 4.11

Min 22.67 17.00 1.00 31.17 5.00 -1.00

Max 144.25 100.00 30.00 352.02 29.00 21.00

(Continued…)

82 No, Kim and Yu

(Table 2 Continued)

Stats SOL RES BID SQU TFL FLO

Northeastern Zone

Mean 91.16 79.81 6.95 83.50 15.94 7.41

S.D. 10.88 9.09 4.11 18.72 3.67 3.07

Min 29.07 27.00 1.00 13.31 5.00 1.00

Max 133.10 100.00 34.00 178.71 27.50 22.00

Southeastern Zone

Mean 87.93 76.94 6.38 106.01 13.33 6.77

S.D. 10.47 8.68 3.70 26.84 3.75 2.60

Min 43.75 41.00 1.00 48.30 5.00 1.00

Max 133.36 100.00 44.00 204.38 27.50 18.00

Northwestern Zone

Mean 87.55 76.22 5.60 93.64 11.98 6.06

S.D. 11.52 10.33 4.16 24.06 3.67 2.84

Min 44.80 33.00 1.00 42.96 5.00 1.00

Max 124.08 100.00 30.00 222.36 24.50 16.00

Southwestern Zone

Mean 89.68 78.58 6.45 87.24 14.60 7.22

S.D. 10.21 9.32 3.65 17.33 4.22 3.49

Min 47.02 41.00 1.00 39.48 5.00 1.00

Max 142.82 100.00 26.00 197.48 28.70 22.00

3.2 Variables

Although researchers, including Nunes and Boatwright (2001), have indicated

that reserve prices influence the WTP of bidders in auctions, the selling prices

in the court auctions of residential real estate are determined by interactions

among numerous factors. Our dependent measure is LSOL (log of rate of selling

price), which is calculated by the ratio of the selling price to the appraised price.

Accordingly, LSOL represents the marginal change in the WTP of bidders in

terms of auction outcomes. We extract 10 independent variables from auction

attributes, building structure, and location conditions that prior researchers

agree serve as the key factors that affect selling prices in housing auctions.

The first category of variables, such as the LRES (log of the rate of the reserve

price)3 and LBID (log of the number of bidders), are auction attributes. The

LRES, the percentage of the difference between the reserve price and the

appraisal price, illustrates the marginal difference in the reference point. In

Korean court auctions, tenants (TENA) and lien holders (LIEN) greatly affect

the transfer prices because special laws that allow the lease and lien of

residential buildings give “opposing power” despite that they are no longer

registered on the record system. In other words, even though the residence is

3 RES (Rate of Reserve Price) = (Reserve Price /Appraisal Price) × 100.

Reserve Prices Yield Reference Price Effects 83

not registered, if the lessee has completed the delivery of a house and the

resident registration, the lease shall take effect against the third person from the

following day thereof (Article 3(1) of the Housing Lease Protection Act) 4 .

Liens are also not registered on the record system. However, a successful bidder

is responsible for reimbursing to the lien holder the claims that are secured by

the lien (Article 91(5) of the Civil Execution Act). For this reason, TENA

(tenants with opposing power) and LIEN (occupants with opposing power) tend

to greatly influence the selling prices in auctions so we have included them as

variables that are auction attributes.

The second category of variables is derived from the characteristics of the

building structure: the LSQU (log of the square meters of the relevant property),

LTFL (log of the total number of floors of an apartment building), and LFLO

(log of the relevant floor of the apartment unit). In Seoul, the traditionally large

family system and continual increase in house prices have established the

tendency to prefer larger dwellings5. The people in the city of Seoul tend to

believe that units on higher floors in apartment buildings mean a higher price.6

They also strongly prefer the royal floors, i.e. the best floor for residence. This

is not a very unique phenomenon. Similarly, Lin (2005) indicates that the house

size, total number of floors, and floor of the residence influence the house prices

both in auction and traditional search markets in Taipei.

The third category of variables is taken from the location conditions: SUBW

(availability of subways), and MART (availability of shopping malls), and

4 Ministry of Government Legislation (2015). Korean Laws in English.

(http://www.moleg.go.kr/english/korLawEng?). 5 According to Statistics Korea (http://www.index.go.kr/potal/), households with four

and more individual comprise about 72.2% of the population in 1980, and 30.6% in 2010,

respectively. The larger family size demands larger sized houses so that people in Seoul

prefer larger sized housing. The continuous increase in house prices has also resulted in

preference for large apartments because these will provide a greater marginal profit. 6 Until the 1990s, the highest residential houses had at most 10-15 stories and the 1/4

rule was accepted for determining the royal floors in Seoul. The traditional ‘1/4 rule’ in

Korea means that the ‘royal floors’ in high-rise housing tend to be defined as the middle

floors except for the 1/4 top and bottom floors of the building respectively. For example,

the ‘royal floors’ in a 20-story apartment are the 6th through to the 15th floors. The ‘royal

floors’ represent the preferred floors of a building by users and investors. Their prices

are usually higher by 5% to 10% than the other floors. The term ‘1/4 rule’ is neither a

legal term nor an official standard criterion. However, it is a widespread common

practice of transactions in Korean high-rise apartment markets up until the 1990s. The

primary reasons for the preference of the middle floors in high rise apartments are

attributed to various complex factors, including noise, privacy, insulation, etc. In other

words, the bottom floors are vulnerable to lack of privacy and the top floors are

disadvantageous due to the heating problem, which at that time with high oil prices, was

a matter of concern because Korea is a non-oil nation. However, in the 2000s, there was

an increase in super high-rise apartments and the royal floors were extended to the higher

floors, which was also due to their abundance of natural light. These super high-rise

apartments were also regarded as a symbol of wealth.

84 No, Kim and Yu

SCHO (availability of schools). Dummy values are then selected, with these

variables taking 1 to represent the availability of schools, subways, and

shopping malls within a vicinity of 1 km or 10 minutes by foot, and 0 otherwise.

In Seoul, the accessibility premiums overwhelmingly offset the negative

amenities from bad externalities such as noise and crime, as proven by

empirical studies on proximity premiums to subway stations and shopping

malls (see Bae et al., 2003; Kim et al., 2005). With 16,500 inhabitants per square

kilometer as of 2003, Seoul is one of the world’s most populated cities. Seoul

is 1.3 times greater in density than Tokyo and twice as much as New York

(Pucher et al., 2005). Accordingly, the reverse or nonlinear problem of

proximity premiums for being right next to subway stations and shopping malls

is not such a decisive factor as far as the analysis of the housing market in Seoul

is concerned 7 . According to 2012 Seoul Statistics (Seoul Metropolitan

Government, 2015), the passing speed of a passenger car is 18.7 km per hour,

and that of a bus is 20.0 km in downtown Seoul. The modal share of daily travel

transportation by subway is 38.2% (12,275 thousand volume of traffic passing

(use)). The traffic congestion and preference for the subway mean that Seoul

citizens accept high housing price premiums for accessibility to subway stations.

The same applies to shopping malls. In Seoul, shopping centers are usually

located near (in) the subway stations and have evolved to become a focal point

for the neighborhoods, aside from retail services.

Contrary to other nations8, entrance exams and school reputation in Seoul do

not pose as serious problems due to the equalization policy and extreme density.

The “equalization policy (pyong-jun-hwa)” and “school district (hak-gun-je)”

of secondary schools have replaced the traditional entrance examinations at

individual high schools for placement of students, and instead, lotteries are used

to randomly assign middle school graduates to high schools within school

districts (Kim, 2004). The average time to travel to school in Seoul is 34.4

minutes. Only 0.6% of the students in Seoul go to school in a private car while

56.5% use public transportation, such as taking the bus or subway. About 31.8%

of the students go to school on foot or by bicycle. For parents and homeowners,

the most important factor that influences house prices is not the equalization

policy or the school quality, but the availability of schools nearby their

residence (Seoul Metropolitan Government, 2014).

7 Studies that have indicated a net negative relationship between house prices and

proximity to subway stations have been conducted in San Diego (Duncan, 2008) and

Chengdu (Cong et al., 2014). However, proximity to rail transit stations can also impose

nuisance effects, like crime and noise to nearby neighborhoods. Some studies have

shown mixed (nonlinear) relationships, such as in Ottawa (Hewitt and Hewitt, 2012).

The same is found for proximity to a shopping mall (see Kim and Zhang, 2005;

Matthews, 2006). 8 Researchers have often applied the hedonic pricing model, in which the house price

premium is assumed constant because the effect of the quality of schools on house prices

is constrained to be linear (Haurin and Brasington, 1996). However, some researchers

have recently proposed the nonlinear effects of school quality from externalities such as

racial composition, crime rate etc. (Chiodo et al., 2010).

Reserve Prices Yield Reference Price Effects 85

3.3 Methods

3.3.1 Panel Data Estimation Model

To examine whether reserve prices influence the final selling prices, we

performed a panel data analysis. The primary motivation for using a panel data

method is to address the problem of omitted variables and examine whether the

unobserved effect is uncorrelated with the explanatory variables (Wooldridge,

2002). A regression model is applied for analysis of the basic panel data as

follows:

it it i ity X β z α ε (1)

There are K regressors initX , not including a constant term. The individual

effect isiz α , where

iz contains a constant term and a set of individual- or

group-specific variables.

Under the assumption of iz , three models are applied to analyze the panel data

as follows: pooled OLS (POLS), fixed effects, and random effects. The POLS

model considers that iz is observed for all individuals and contains only a

constant term. Thus, the entire OLS model can sufficiently estimate both the

common α and the slope vector β . The fixed effects model assumes that iz is

unobserved but correlated with itX . Therefore, the OLS estimator of β is

biased and cannot provide consistent estimates. In this model, iz α (=

iα )

represents all of the observable effects and taken to be a group-specific constant

term. Finally, the random effects model postulates that the unobserved

individual heterogeneity can be considered to be uncorrelated with the included

variables.

The quality of a property and the location tend to exhibit highly autoregressive

correlations because of spatial dependence and heterogeneity, while there is

also a relationship between housing prices and unobservable location factors

(Tse, 2002). In addition, housing prices evolve over time because the housing

market is dynamic. Thus, we compare the statistical results of the POLS and

fixed effects models with a double log form in order to estimate the panel data

regression. In contrast to the POLS model, two unobservable variables that

exemplify individual effects (iγ ) and time effects (

tγ ) are added into the fixed

effects model. Our POLS (fixed) model can thus be expressed as follows:

0 1 2 3 4 5

6 7 8 9 10

( ) ( )

it it it it it it

it it it it it

i t it

LSOL β β LRES β LBID β LSQU β LTFL β LFLO

β SCHO β SUBW β MART β TENA β LIEN

γ γ ε

(2)

where

Assets (i): Individual apartments

86 No, Kim and Yu

Period (t): Time (years)

Rate of selling (LSOLit): Log of rate of selling price to appraised price (%)

Reserve price rate (LRESit): Log of rate of reserve price to appraised price (%)

Bidders (LBIDit): Log of the number of participating bidders (persons)

Square (LSQUit): Log of the total square meters of an apartment (m2)

Total floor (LTFLit): Log of the total number of floors in an apartment building

(floors)

Floor (LFLOit): Log of the relative floor of an apartment (floor)

School (SCHOit): School available (1); otherwise (0)

Subway (SUBWit): Subway available (1); otherwise (0)

Mart (MARTit): Shopping mall available (1); otherwise (0)

Tenant (TENAit): Opposing tenants (1); otherwise (0)

Lien (LIENit): Opposing occupants (1); otherwise (0)

1 2 3 10, , , ,β β β β : Coefficients of the independent and control variables

iγ : Characteristics of an individual asset

tγ : Characteristics of time

itε : Remaining stochastic disturbance terms which are interpreted as the

effect of unspecified variables of the characteristics of individual assets

and time

3.3.2 Double LOLS Model

We introduce the LOLS9 model to examine how the reference price effects of

reserve prices differ according to housing size. We transform a simple OLS

model into a double LOLS model to remedy the possible violation of the normal

distribution assumption on the unobserved error terms (Hardin and Wolverton,

1996). In the case of the housing market, a double log function is preferred to a

linear function because the former measures the elasticity of each characteristic,

while the WTP for an additional unit of the characteristic is not fixed but rather

decreases. The specification of the double log model in this study used to assess

the sensitivity of reference price effects is

0 1 2 3 4 5

6 7 8 9 10

j j j j j j

j j j j j j

LSOL β β LRES β LBID β LSQU β LTFL β LFLO

β SCHO β SUBW β MART β TENA β LIEN ε

(3)

3.3.3 RSM Model

9 We use the OLS regression model in lieu of a panel regression in the analysis of the

sensitivity of the reference price effects of reserve prices. A panel estimation is

unsuitable in this case because of its lopsided sample size and the time characteristics

when our data are classified in accordance with the scale and period of time.

Reserve Prices Yield Reference Price Effects 87

We use the RSM to predict the expected rate of the selling price in accordance

with fluctuations in the reserve prices. RSM is defined as a collection of

mathematical and statistical methods that are used to develop, improve, or

optimize a product or process (Myers and Brenneman, 2010). Accordingly,

RSM is typically used to locate the optimum response of a dependent variable

or predict how the dependent variable changes in a given direction by

conjecturing the functional relation from the data. The relation between price

and the characteristics of housing properties is likely to be nonlinear. Thus, a

nonparametric estimation analysis such as the RSM allows us to effectively

estimate this nonlinearity.

The true RSM form of the function ∫ is unknown because the relationship

between the dependent and independent variables is usually unknown. The

RSM is widely used when input factors may affect the outputs; thus, the former

is often called the response, while the latter are the independent variables. In

most statistical analyses, the RSM function is formulated by using coded

variables, namely 1x ,

2x , , and kx which are typically defined to be

dimensionless with a mean zero and the same standard deviation (Carley,

Kamneva, and Reminga, 2004). The RSM function with coded variables can

thus be represented as

1 2, , , kη f x x x (4)

A fundamental method for assessing quantitative variables involves the fitting

of the first-order or second-order functions of the predictors to one or more

response variables and then examining the characteristics of the fitted surface

to decide what action is appropriate (Lenth, 2009). The first-order model

approximation of the function ∫ is suitable under the condition that it is not too

curved. Therefore, a polynomial model is usually sufficient and the first-order

model is assumed to be an adequate approximation of the true surface in a small

region of xs (Bradley, 2007). However, the first-order model is not sufficient

when the curvature in the response surface is too strong, meaning that a second-

order-model is likely to be required (Carley, Kamneva, and Reminga, 2004). In

general, the second-order model can be expressed as (Montgomery, 2005):

12

0

1 1 1 1

1, 2, ,n n n n

i i ii i ij i j

i i i j i

Y β β x β x β x x i n

(5)

where Y is the predicted response by the model, 0β is a constant,

iβ is the

linear coefficient, iiβ the second-order coefficient, ijβ the interaction

coefficient, and ix and jx are the coded values that correspond to the tested

variables (Kalali et al., 2011).

88 No, Kim and Yu

In this study, we use the second-order model because of its distinct advantages10.

Thus, our fundamental RSM model can be expressed as follows:

2 2

1 1 2 2 3 1 4 2 1 5 2i i iY α β X β X β X β X X β X ε (6)

where Yi = rate of sold price, iα = constant,

1X = rate of reserve price, 2X =

number of bidders, and iε = error term. This second-order RSM model is

useful for approximating a proportion of the true response surface by using a

parabolic curvature. It includes all the terms in the first-order model as well as

all the quadratic terms (e.g., 2

3 1β X ) and cross-product terms (e.g., 4 2 1β X X )

(Bradley, 2007).

In our model, the coefficient 1β represents the slope of the linear response

curve to the rate of the reserve price (1X ) and

2β is the slope of the linear

response curve to the number of bidders (2X ). The coefficients

3β and 5β

represent the response curvatures that are related to 2

1X and 2

2X , respectively,

while 4β expresses the synergy effect of the interaction between

2X and 1X

(Bettinger and Chinnici, 1991). For simplicity and clarity, we have omitted 1X

and 2X from the presented analyses.

4. Empirical Results

4.1 Reference Price Effects of Reserve Prices

For the panel data estimates, we must perform a unit root test to verify the

stationarity of the time-series data. The results of the unit root test for the SMA

panel data reject the null hypothesis of all the panels that contain unit roots at

the 1% significance level (Table 3), that is, the panel data are stationary. Before

the panel data analysis, we must also perform the Hausman test to select the

estimation method for the fixed and random effects models. The Hausman test

results presented in Table 4 show that the initial hypothesis, which states that

the random effects model is adequate, is rejected at the 1% significance level.

Based on the results of the Hausman test, we examine the panel data by using

the fixed effects model.

10 The second-order model is widely used in RSM for several reasons. First, its flexibility

allows the model to take on a wide variety of functional forms, which means that it will

often work well as an approximation of the true response surface. Second, it is easy to

estimate the parameters (the βs) in the second-order model by using the method of least

squares. Finally, considerable practical experience indicates that second-order models

work well in solving real response surface problems (Carley, Kamneva, and Reminga,

2004).

Reserve Prices Yield Reference Price Effects 89

Table 3 Unit Root Test for the Panel Data

LSOL LRES LBID LSQU LTFL LFLO

Inverse x2 485.2549*** 502.6076*** 845.3829*** 911.0499*** 939.7198*** 950.4041***

Inverse normal -18.5014*** -19.0732*** -25.9423*** -27.5187*** -27.9919*** -28.0786***

Inverse logit t -27.9769*** -29.0037*** -48.7983*** -52.5904*** -54.2454*** -54.8621***

Modified inv. x2 45.7955*** 47.6046*** 83.3414*** 90.1877*** 93.1767*** 94.2906***

Note: 1) We perform Fisher-type unit root tests for the unbalanced panel data based on augmented Dickey–Fuller tests.

2) H0: All panels contain unit roots, Ha: At least one panel is stationary.

3) ***, **, and * represent a significance levels of 1%, 5%, and 10%, respectively.

Reserv

e Prices Y

ield R

eference P

rice Effects 8

9

90 No, Kim and Yu

Table 4 Hausman Test for the Panel Data

Variable

Fixed

Effects

(βс)

Random

Effects

(βe)

Difference

(βс- βe)

S.E.

sqrt(diag(Vс-Ve))

LRES 0.905065 0.905136 -0.0000708 0.0015869

LBID 0.054339 0.0525 0.0018391 0.0004041

LSQU -0.01285 -0.01159 -0.0012558 0.0029247

LTFL -0.02609 -0.03154 0.0054507 0.0020532

LFLO -0.00103 -0.0011 0.0000685 0.0003239

SCHO -0.00696 -0.00991 0.0029521 0.0008442

SUBW 0.000337 0.001204 -0.0008676 0.0011871

MART -0.01566 -0.00667 -0.0089906 0.0029947

TENA -0.00868 -0.00998 0.0012971 0.0004474

LIEN -0.00244 -0.00334 0.0008987 0.0006243

x2 (2) = (βс- βe)’ (Vс-Ve)-1 (βс- βe) = 37.69

Prob > x2 0.0000

Note: 1) We perform the Hausman test for the panel data by using the Seoul metropolis

as a whole

2) βc is the coefficient vector from the consistent estimator

3) βe is the coefficient vector from the efficient estimator

4) Vc is the covariance matrix of the consistent estimator

5) Ve is the covariance matrix of the efficient estimator

6) H0: difference in coefficients not systematic

We report the parameter estimates for the POLS and unbalanced panel

regression with fixed effects in Table 5. We find that LRES (rate of the reserve

price) and LBID (number of bidders) exert a meaningful positive influence onto

LSOL (rate of the selling price) at the 1% significance level. On the contrary,

LSQU (size of the apartment), LTFL (total number of floors of the apartment

building), SCHO (school availability), MART (shopping mall availability), and

TENA (tenants with opposing power) have a significant negative effect on the

LSOL in both models. However, the remaining independent variables, the LFLO

(relevant floor of the apartment), SUBW (subway availability), and LIEN

(occupants with opposing power) do not have a significant effect on the LSOL

at the same level of statistical significance.

The influence of a certain independent variable on the dependent variable in a

double log regression model reflects its elasticity under the control of the

remaining independent variables. Consequently, the coefficients 0.9052 and

0.9051 of LRES in the outcomes of the POLS and fixed effects models indicate

the existence of reference price effects. In other words, if the district courts are

to increase the reserve price by 1%, the selling price would increase by 0.9052%

with the POLS model and by 0.9051% with the fixed effects model (Rosen,

1974).

Reserve Prices Yield Reference Price Effects 91

Table 5 Parameter Estimates for the OLS and Panel Regression

Variable

SMA Central Zone Northeastern Zone Southeastern Zone Northwestern Zone Southwestern Zone

POLS Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects

LRES 0.9052

(91.38)***

0.9051

(91.14)***

0.9353

(29.99)***

0.9332

(29.16)***

0.9551

(52.25)***

0.9542

(51.62)***

0.9449

(43.72)***

0.9453

(43.49)***

0.8563

(31.82)***

0.8605

(32.30)***

0.8619

(47.32)***

0.8571

(46.52)***

LBID 0.0524

(26.71)***

0.0543

(27.38)***

0.0707

(10.77)***

0.0715

(10.83)***

0.0603

(17.87)***

0.0605

(17.65)***

0.069

(15.20)***

0.0685

(15.05)***

0.0463

(9.05)***

0.0446

(8.70)***

0.0346

(9.38)***

0.0347

(9.36)***

LSQU -0.0116

(-2.29)**

-0.0128

(-2.21)**

-0.0248

(-1.57)

-0.0324

(-1.85)*

-0.0141

(-1.50)

-0.0127

(-1.22)

-0.0122

(-1.17)

-0.0157

(-1.28)

0.016

(-0.95)

0.0308

(1.74)*

-0.018

(-1.51)

-0.0177

(-1.47)

LTFL -0.0317

(-6.94)***

-0.0261

(-5.25)***

-0.0211

(-1.26)

-0.0204

(-1.21)

-0.0155

(-1.66)*

-0.0174

(-1.80)*

-0.0299

(-2.95)***

-0.0284

(-2.74)***

-0.0242

(-1.69)*

-0.018

(-1.25)

-0.033

(-3.99)***

-0.0331

(-3.79)***

LFLO -0.0011

(-0.40)

-0.001

(-0.38)

0.01

(-1.09)

0.0103

(-1.11)

-0.0199

(-4.53)***

-0.0194

(-4.37)***

-0.0034

(-0.52)

-0.004

(-0.60)

-0.0019

(-0.22)

-0.0024

(-0.29)

0.0119

(2.53)**

0.0115

(2.44)**

SCHO -0.01

(-1.81)*

-0.007

(-1.26)

-0.0142

(-0.78)

-0.0159

(-0.87)

0.0343

(1.82)*

0.0353

(1.85)*

0.0088

(-0.54)

0.0089

(-0.54)

-0.0064

(-0.51)

-0.0104

(-0.82)

-0.0104

(-1.32)

-0.009

(-1.13)

SUBW 0.0012

(-0.27)

0.0003

(-0.07)

-0.0143

(-0.82)

-0.0172

(-0.96)

-0.0258

(-1.71)

-0.026

(-1.68)*

0.0493

(3.01)***

0.0478

(2.91)***

0.018

(1.78)*

0.0144

(-1.43)

-0.002

(-0.33)

-0.0016

(-0.26)

(Continued…)

Reserv

e Prices Y

ield R

eference P

rice Effects 9

1

92 No, Kim and Yu

(Table 5 Continued)

Variable

SMA Central Zone Northeastern Zone Southeastern Zone Northwestern Zone Southwestern Zone

POLS Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects POLS

Fixed

Effects

MART -0.0065

(-2.33)**

-0.0157

(-3.86)***

-0.0238

(-1.69)*

-0.0235

(-1.66)*

-0.0342

(-3.54)***

-0.035

(-3.54)***

-0.0408

(-4.29)***

-0.0433

(-4.51)***

-0.0008

(-0.09)

-0.0163

(-1.53)

-0.0041

(-0.74)

-0.0053

(-0.81)

TENA -0.01

(-3.89)***

-0.0087

(-3.35)***

-0.0049

(-0.41)

-0.0069

(-0.57)

-0.0055

(-1.25)

-0.0063

(-1.41)

-0.0145

(-3.02)***

-0.0145

(-3.03)***

-0.0098

(-1.32)

-0.0055

(-0.73)

-0.003

(-0.70)

-0.0019

(-0.45)

LIEN -0.0034

(-1.05)

-0.0024

(-0.75)

0.0041

(-0.23)

0.0046

(-0.26)

0.0004

(-0.07)

-0.0005

(-0.08)

-0.0028

(-0.51)

-0.0023

(-0.41)

-0.0054

(-0.65)

-0.0044

(-0.53)

0.0005

(-0.09)

0.0001

(-0.02)

Cons 0.6142

(11.66)***

0.6046

(11.03)***

0.5138

(3.09)***

0.5578

(3.09)***

0.39

(4.13)***

0.3927

(4.04)***

0.381

(3.16)***

0.3972

(3.13)***

0.679

(4.37)***

0.5901

(3.71)***

0.8341

(8.46)***

0.8526

(8.49)***

Adj. R2 /

overall

R2

0.8768 0.8765 0.8951 0.9021 0.8976 0.9003 0.9033 0.9068 0.8819 0.8845 0.8529 0.8562

F / Wald 1010.49*** 986.82*** 125.61*** 123.11*** 325.42*** 308.43*** 251.23*** 247.62*** 139.93*** 141.62*** 258.43*** 245.71***

Note: The dependent variable is the log rate of the selling price to the appraised price (SOL).

92

No, K

im an

d Y

u

Reserve Prices Yield Reference Price Effects 93

The zonal results show that LRES and LBID have strong positive effects on the

rate of the selling price (LSOL) at the 1% significance level in all five zones of

the SMA. However, the parameter estimates differ by zone. For the POLS

model, the largest coefficient of LRES is 0.9551 (northeastern), while the

weakest is 0.8563 (northwestern); for the fixed effects model, the strongest

coefficient of LRES is 0.9542 (northeastern) and the weakest is 0.8571

(southwestern). The remaining independent and dummy variables (LSQU,

LTFL, LFLO, SCHO, SUBW, MART, TENA, and LIEN) show inconsistent signs

(+ or –) and statistical significances according to the zone under study.

The panel regression estimates show that the basic reference price effects of

reserve prices exist in housing auctions both across the SMA as a whole and in

its five zones. We also find that the coefficients of both the POLS and the fixed

effects models are almost identical. Therefore, by analyzing the reference price

effects with the use of the panel data, we infer that the reference price effects of

LRES on LSOL are very strong even though regional differences exist.

4.2 Sensitivity of Reference Price Effects

According to the seminal prospect theory, the marginal value of both gains and

losses in monetary changes generally decreases with their magnitude. Contrary

to the expectations of utility theorists, behavioral economics theorists suggest

that value should thus be treated as a function of the asset position that serves

as the reference point and the magnitude of the change from that reference point

(Kahneman and Tversky, 1979). In practice, the perceived price is greater when

for example, the price of a vehicle increases from $10,000 to $11,000 compared

to $100,000 to $101,000 (i.e., the same difference of $1000). Against this

background, we perform an LOLS regression to investigate whether the

sensitivity of the reference price effects differs with housing size. If the

arguments of the behavioral economics theorists are sound, we may find greater

sensitivity of the reference price effects of reserve prices for smaller apartments

compared to the larger ones.

Table 6 summarizes the parameter estimates for the sensitivity analysis with the

LOLS regression. The parameter estimates of LRES are positive, thus implying

that reserve prices influence selling prices in all apartment size groups. In other

words, there are consistently reference price effects on the reserve prices

regardless of the apartment size. On the contrary, the sensitivity of the reference

price effects derived from the LOLS regression is somewhat different. Among

the six apartment size groups, the reference price effects are the strongest for

the 111–132 m2 apartments and weakest for the 133–164 m2 apartments. The

results of the sensitivity analysis, however, do not provide coherent information

on the reference price effects of the reserve prices.

94 No, Kim and Yu

Table 6 Parameter Estimates for Sensitivity Analysis with OLS Regression (SMA)

Variable Under 60 m2 61–84 m2 85–110 m2 111–132 m2 133–164 m2 Over 165 m2

18 pyeong type 25 pyeong type 33 pyeong type 40 pyeong type 50 pyeong type Over 60 pyeong

LRES 0.900(29.19)*** 0.938(52.35)*** 0.867(46.50)*** 0.98432.95)*** 0.828(11.95)*** 0.921(12.92)***

LBID 0.065(7.29)*** 0.057(18.86)*** 0.041(12.85)*** 0.047(8.67)*** 0.056(6.04)*** 0.073(2.92)***

LSQU 0.032(-0.82) 0.004(-0.19) 0.003(-0.15) -0.026(-0.44) 0.095(-0.77) 0.032(-0.3)

LTFL -0.074(-3.83)*** -0.020(-2.78)*** -0.040(-5.36)*** -0.031(-2.56)** 0.004(-0.17) -0.015(-0.27)

LFLO -0.012(-1.08) -0.003(-0.77) 0.000(-0.01) -0.000(-0.05) -0.019(-1.27) 0.034(-1.26)

SCHO 0.016(-0.77) -0.025(-2.83)*** -0.018(-1.79)* -0.005(-0.34) -0.022(-0.57) 0.053 (-1.37)

SUBW -0.006(-0.31) -0.001(-0.20) -0.000(-0.05) 0.020(1.72)* 0.032(-1.11) 0.016(-0.35)

MART 0.037(1.75)* -0.005(-1.14) -0.001(-0.36) -0.022(-2.98)*** -0.031(-1.37) -0.052(-1.51)

TENA -0.013(-0.70) -0.009(-2.20)** -0.006(-1.83)* -0.003(-0.47) -0.018(-1.30) -0.037(-0.89)

LIEN 0.048(1.68)* -0.003(-0.53) -0.003(-0.86) 0.001(-0.22) -0.016(-0.90) 0.004(-0.12)

Constant 0.536(2.63)** 0.382(2.92)*** 0.752(5.15)*** 0.330(-1.05) 0.349(-0.44) 0.164(-0.24)

Sample 82 581 514 157 56 29

Adj. R2 0.928 0.8409 0.824 0.8857 0.8893 0.9001

F-stat 105.44*** 307.56*** 241.25*** 121.91*** 45.20*** 26.22***

Note: 1) The dependent variable is the log rate of the selling price to the appraised price (LSOL).

2) ( ) implies t-statistics and ***, **, and * represent a significance level of 1%, 5%, and 10%, respectively.

3) Pyeong is a unit in Korea that is used to measure the size of the floor space and land (1 pyeong = 3.3 m2 (1.818 m × 1.818 m).

9

4 N

o, K

im an

d Y

u

Reserve Prices Yield Reference Price Effects 95

4.3 Predicting Expected Reference Price Effects

In Korean court auctions of residential real estate, the rate of the selling price

(SOL) is the most important indicator of the likely price of similar properties in

the neighboring area. Bidders thus decide their bids based on their own

estimations of the expected rate of the selling price, which dictate the success

of their bidding. Hence, in this study, a statistical technique is proposed for

predicting the rate of the selling price, namely, a nonparametric RSM. By using

panel data regression and sensitivity analysis, we find that the reserve price

(RES) and number of bidders (BID) strongly influence the rate of the selling

price (SOL) compared to other factors. For simplicity, we have carried out the

RSM analysis by using RES and BID as the independent variables and SOL as

the dependent variable to approximate the expected rate of the selling price (i.e.,

reference price effects) in accordance with the changes in the rate of the reserve

price.

Table 7 presents the parameter estimates for the RSM. In the SMA, all the

parameter estimates of the Xs (1X ,

2X ,1 1X X ,

2 1X X ,2 2X X ) are significant at

the 1% level. However, the magnitude and direction of these estimates are

different. The rate of the reserve price (1X ) has a significant positive effect

(+0.6724) on the rate of the selling price (iY ), whereas the number of bidders

(2X ) have a negative influence (-0.6085) on

iY . The parameter estimate of

2 1X X also indicates the existence of a positive synergistic effect (+0.0200) of

the interactions between the rate of the reserve price (1X ) and the number of

bidders (2X ) over the rate of the selling price (

iY ).

Table 8 shows the expected rates of the selling prices. These results could be very

useful for formulating an appropriate auction strategy. For example, auctioneers

(courts) could easily find a reasonable approximation of the RES without repeated

auction rounds of conflicting prices while bidders may be able to decide their

entry point more objectively in order to avoid the winner’s curse12. Figure 3

presents the contours of the RSM estimations, which show dynamic stimulus–

response relationships between the rate of the reserve price (RES) and the rate of

the selling price (SOL). The highly irregular curvatures and intervals of the

contours suggest that the reference price effects of reserve prices differ by locality

and reserve price.

12 The winner’s curse is where the winning bidders tend to overpay compared with the

objective value of auctioned properties. If this happens, the winner of the auction is

likely to be a loser. The winner can be said to be ‘cursed’ in one of two ways: (1) the

winning bid exceeds the value of the tract, so the firm loses money or (2) the value of

the tract is less than the expert’s estimate, meaning that the winning firm is disappointed.

However, because the winner’s curse cannot occur if all bidders are rational, evidence

in market settings would constitute an anomaly. The winner’s curse is a problem that is

amenable to investigation by using modern behavioral economics, a combination of

cognitive psychology and microeconomics (Thaler, 1988).

96 No, Kim and Yu

Table 7 Parameter Estimates for RSM

Area Parameter estimate

Intercept X1(RES) X2(BID) X1(RES)*X1(RES) X2(BID)*X2(BID) X1(RES)*X2(BID)

SMA 21.012

(5.86)***

0.672

(7.28)***

-0.608

(-2.79)***

0.001

(2.72)***

-0.009

(-3.43)***

0.020

(7.55)***

Central Zone 12.190

(-1.32)

0.740

(2.98)***

0.177

(-0.24)

0.001

(-1.07)

-0.044

(-3.54)***

0.023

(2.96)***

Northeastern Zone 18.321

(2.57)**

0.735

(4.07)***

-0.500

(-1.23)

0.001

(-0.83)

-0.029

(-5.79)***

0.023

(4.99)***

Southeastern Zone 27.204

(3.11)***

0.427

(1.87)*

-0.507

(-0.90)

0.003

(2.30)**

-0.003

(-1.13)

0.020

(2.84)***

Northwestern Zone 21.709

(2.51)**

0.600

(2.60)**

0.497

(-0.97)

0.002

(-1.59)

-0.025

(-2.85)***

0.008

(-1.33)

Southwestern Zone 26.661

(3.30)***

0.597

(2.99)***

-0.69

(-1.52)

0.002

(-1.55)

-0.004

(-0.54)

0.016

(3.17)***

Note: 1) The dependent variable is the rate of the selling price to the appraised price (SOL).

2) ( ) implies t-statistics and ***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively.

9

6 N

o, K

im an

d Y

u

Reserve Prices Yield Reference Price Effects 97

Table 8 Prediction of the Expected Rates of Selling Price

Area

Rate of

reserve price

(RES: %)

Average number

of bidders

(BID: persons)

Expected rate of

selling price

(SOL: %)

SMA

100 5.68 112.862

80 5.844 91.168

72 6.727 83.42

64 7.132 75.335

51 4.7 61.459

Central Zone

100 5.906 117.557

80 5.117 92.278

72 7.4 86.187

64 6.475 75.943

51 3 58.361

Northeastern

Zone

100 5.333 111.086

80 6.038 90.994

72 6.909 83.475

64 6.991 75.236

51 3 60.309

Southeastern

Zone

100 0.75 106.129

80 6.957 91.514

72 6.071 81.881

64 7.94 75.06

51 5 60.715

Northwestern

Zone

100 6.5 114.388

80 5.915 91.785

72 6.542 84.041

64 7.048 76.43

51 1 59.733

Southwestern

Zone

100 5.135 111.458

80 5.76 90.954

72 6.505 83.317

64 7.713 75.882

51 6.667 63.31

Note: The expected rate of the selling price (SOL) is calculated by using estimated RSM

functions.

98 No, Kim and Yu

Figure 3 Contours of the RSM Estimation

Note: 1) The horizontal axis is the rate of the reserve price (%).

2) The vertical axis is the number of bidders.

3) The contour is the rate of the selling price for the SMA and the five zones of

the SMA.

5. Conclusion

Recently, a number of theoretical auction studies have acknowledged that

reserve prices serve as a reference point which may influence the WTP of

bidders. However, both the theoretical and the empirical results of research

work that explored the reference price effects of reserve prices in auctions are

inconclusive (Trautmann and Traxler, 2010). Nevertheless, the reserve prices

created by the courts substantially affect the interests of the parties concerned.

By reducing the reserve prices, for example, the courts can shorten the bid

timings and avoid breaking down of tenders (Lin, 2005). However, extremely

low selling prices would reduce the benefits for both creditors and debtors. On

the contrary, higher reserve prices may inhibit the possibility of a sale and result

in repeated auctions (McAfee, Quan, and Vincent, 2002).

Reserve Prices Yield Reference Price Effects 99

Based on the foregoing, the effects of reserve prices in court auctions of

residential real estate in Korea have been examined in this study. The results of

the panel data estimates presented herein show that the reserve price level has

significant reference price effects on transfer prices, thus suggesting that there

are strong reference price effects in the court auctions of residential real estate

of the SMA as a whole as well as in its five zones. However, the magnitude of

the reference price effects of reserve prices differs by zone. Furthermore, we

find that the sensitivity of the reference price effects derived from the LOLS

regression differs with housing size. Among the six groups of apartment size

analyzed, reference price effects are shown to be the strongest for the 111–132

m2 apartments and weakest for the 133–164 m2 apartments. The results of the

sensitivity analysis, therefore, are not able to provide coherent information on

the reference price effects of the reserve prices.

Finally, the prediction analysis which uses a nonparametric RSM proposes that

varying reserve prices leads to differences in the expected reference price

effects with locality. Therefore, courts and bidders may be able to approximate

the expected rate of the selling price by using the RSM simulations described

in this study, which would allow them to set reserve prices and decide on the

bid levels.

Resultant of the study findings, some policy suggestions are recommended to

improve the court auction process of residential real estate in Korea. First,

current prices should be appropriately appraised and reserve prices reasonably

stipulated because these control the outcome of auctions through the reference

price effects described herein. During the period of this study, only one-third

(32.7%) of the court auctions of real estate examined are successfully carried

out. The low probability of selling (i.e., high degree of discrepancy between

reserve and reference prices) suggests that the reserve prices are not being

established at a reasonable amount by appraisers and/or the courts. Second, the

cutoff intervals (20–30%) of the reserve prices in cases of discrepancies

between the reserve and reference prices must be reduced (e.g., to 5–10%).

Such a wide gap between auction rounds means that there is susceptibility to

bias with the drawing of artificial reference points. For instance, it is suggested

that Korean tax offices reduce reserve prices by 10% when public sales fail in

delinquency dispositions.

Finally, some future directions for research are provided. First, researchers can

carry out more studies on real estate auction markets. In the course of

performing this study, we found that the literature on real estate auctions is still

lacking despite the numerous available sources. Second, there should be a

balance in the research on first-price sealed-bid auctions. Previous works on

reserve prices have focused on auctions in the western context; thus, increasing

the scope of the research on sealed-bid auctions would be desirable to provide

researchers with generalized theories and practices in other countries that have

adopted the first-price sealed-bid auction system (e.g., Taiwan, Nigeria,

Singapore, and the US).

100 No, Kim and Yu

Acknowledgement

This work was supported by the National Research Foundation of Korea Grant

funded by the Korean Government (NRF-2012S1A3A2033330).

References

Ariely, D. and Simonson, I. (2003). Buying, Bidding, Playing, or Competing?

Value Assessment and Decision Dynamics in Online Auctions, Journal of

Consumer Psychology, 13, 1&2, 113–123.

Bae, C.H.C., Jun, M.J. and Park, H. (2003). The Impact of Seoul's Subway Line

5 on Residential Property Values, Transport Policy, 10, 2, 85-94.

Bae, H.J. (2013) Korean Real Estate Market Highly Attractive for Investors,

The Korea Herald, July 16, 2013. Available at: http://www.koreaherald.com.

Bettinger, D. and Chinnici, J.P. (1991). Utilization of Response Surface

Modeling to Evaluate Interaction between Aflatoxin B1 and Caffeine on Egg-

Adult Viability in Drosophila Melanogaster, Virginia Journal of Science, 42,

311–320.

Bourassa, S., Hamelink, F., Hoesli, M. and MacGregor, B. (1999). Defining

housing submarkets, Journal of Housing Economics, 8, 2, 160–183.

Bradley, N. (2007). The Response Surface Methodology, Master’s dissertation,

Indiana University of South Bend.

Carley, K.M., Kamneva, N.Y., and Reminga, J. (2004). Response Surface

Methodology, CASOS Technical Report, CMU-ISRI-04-136, 1–26.

Chiodo, A.J., Hernández-Murillo, R., and Owyang, M.T. (2010). Nonlinear

Effects of School Quality on House Prices. Federal Reserve Bank of St. Louis

Review, 92, 3, 185-204.

Cong, S., Zheng, S., and Han, R. (2014). Proceedings of the 17th International

Symposium on Advancement of Construction Management and Real Estate:

The Effects of Subway Construction on Housing Premium: A Macro-data

Reserve Prices Yield Reference Price Effects 101

Analysis in Chengdu’s Housing Market, ISBN, 78-3-642-35548-6, Springer-

Verlag, Berlin Heidelberg, 445-454.

Dewally, M. and Louis Ederington, L. (2004). What Attracts Bidders to Online

Auctions and What Is Their Incremental Price Impact? Available at SSRN:

http://ssrn.com/abstract=589861 or http://dx.doi.org/10.2139/ssrn.589861.

Duncan, M. (2008). Comparing Rail Transit Capitalization Benefits for Single-

Family and Condominium Units in San Diego, California, Transportation

Research Record, 2067, 120–130.

Hardin, W. and Wolverton, M. (1996). The Relationship between Foreclosure

Status and Apartment Price, Journal of Real Estate Research, 12, 1, 101–109.

Haurin, D.R. and Brasington, D. (1996). School Quality and Real House Prices:

Inter-and Intra-meTropolitan Effects, Journal of Housing Economics, 5, 351-

368.

Hewitt, C.M. and Hewitt, W.E. (2012). The Effect of Proximity to Urban Rail

on Housing Prices in Ottawa, Journal of Public Transportation, 15, 4, 43–65.

Kamins, M., Dréze, X., and Folkes, V. (2004). Effects of Seller Supplied Prices

on Buyers’ Product Evaluations: Reference Prices in An Internet Auction On

Text, Journal of Consumer Research, 30, 4, 622–628.

Kahneman, D. and Tversky, A. (1979). Prospect theory: An Analysis of

Decision under Risk, Econometrica, 47, 2, 263–292.

Kalali, A., Ebadi, T., Rabbani, A., and Moghaddam, S. (2011). Response

Surface Methodology Approach to the Optimization of oil Hydrocarbon

Polluted Soil Remediation Using Enhanced Soil Washing, International

Journal of Environmental Science and Technology, 8, 2, 389–400.

Kim, J. and Zhang, M. (2005). Determining Transit’s Impact on Seoul

Commercial Land Values: An Application of Spatial Econometrics,

International Real Estate Review, 8, 1, 1-26.

Kim, T. (2004). Shadow Education: School Quality and Demand for Private

Tutoring in Korea, KDI School of Pub Policy & Management Paper No. 04-21,

1-22. Available at SSRN: http://ssrn.com/abstract=635864

Kristensen, H. and Gärling, T. (1997). Anchor Points, Reference Points, and

Counter Offers in Negotiation, Göteborg Psychological Reports, 27, 1–11.

Ku, G., Galinsky, A., and Murnighan, K. (2006). Starting Low But Ending High:

A Reversal of the Anchoring Effect in Auctions, Journal of Personality and

Social Psychology, 90, 6, 975–986.

102 No, Kim and Yu

Lenth, R. (2009). Response-Surface Methods in R, Using RSM, Journal of

Statistical Software, 32, 7, 1–17.

Lin, V. (2005). Hedonic Model of Court Auction Residential Housing Market

in the Taipei Metropolitan Area, The Queensland University of Technology

Research Week 2005 Conference Proceedings, 1–19.

Matthews, W. (2006). The Effect of Proximity to Commercial Uses on

Residential Prices, Ph.D. Dissertation, Georgia State University and Georgia

Institute of Technology, 1-188.

Mayer, C. (1998). Assessing the Performance of Real Estate Auctions, Real

Estate Economics, 26, 1, 41–66.

McAfee, P., Quan, D., and Vincent, D. (2002). How to Set Minimum

Acceptable Bids, with An Application to Real Estate Auctions, Journal of

Industrial Economics, 50, 4, 391–416.

Milgrom, P. and Weber, R. (1982). A Theory of Auctions and Competitive

Bidding, Econometrica, 50, 5, 1089–1122.

Montgomery, D. (2005). Design and Analysis of Experiments: Response

Surface Method and Designs, New Jersey, John Wiley and Sons.

Myers, W. and Brenneman, W. (2010). Encyclopedia of Biopharmaceutical

Statistics, 3rd ed., London, UK, Information Healthcare.

Nunes, J. and Boatwright, P. (2001). Incidental Prices and Their Effect on

Consumer Willingness to Pay, Working Paper, 1–47.

Popkowski Leszczyc, P., Qiu, C., and He, Y.F. (2009). Empirical Testing of the

Reference-Price Effect of Buy-Now Prices in Internet Auctions, Journal of

Retailing, 85, 2, 211–221.

Pucher, J., Park, H.Y., Kim, M.H., and Song, J.M. (2005). Public Transport in

Seoul: Meeting the Burgeoning Travel Demands of A Megacity, Public

Transport International, 54, 3, 54-61.

Reiley, D.H. (2005). Field Experiments on the Effects of Reserve Prices in

Auctions: More Magic on the Internet, The RAND Journal of Economics, 37, 1,

195-211.

Rosen, S. (1974). Hedonic Price and Implicit Markets: Product Differentiation

in Pure Competition, The Journal of Political Economy, 82, 1, 34–55.

Reserve Prices Yield Reference Price Effects 103

Rosenkranz, S. and Schmitz, P. (2007). Reserve Prices in Auctions as Reference

Points, Economic Journal, 117, 520, 637–653.

Thaler, R. (1988). The Winner’s Curse, Journal of Economic Perspectives, 2, 1,

191–202.

Trautmann, S. and Traxler, C. (2010). Reserve Prices as Reference Points –

Evidence from Auctions for Football Players at Hattrick.org, Journal of

Economic Psychology, 31, 2, 230–240.

Tse, R. (2002). Estimating Neighborhood Effects in House Prices: Towards a

New Hedonic Model Approach, Urban Studies, 39, 7, 1165–1180.

Vincent, D. (1995). Bidding Off the Wall: Why Reserve Prices May Be Kept

Secret, Journal of Economic Theory, 65, 2, 575–584.

Walley, M. and Fortin, D. (2005). Behavioral Outcomes from Online Auctions:

Reserve Price, Reserve Disclosure, and Initial Bidding Influences in the

Decision Process, Journal of Business Research, 58, 10, 1409–1418.

Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data,

Cambridge, Mass., The MIT Press.

Attorneys at Law YULCHON (2012). How the Auction Procedure Works in

Korea, Real Estate And Finance Newsletter, 1-2. http://www.yulchon.com)

Ministry of Government Legislation (2015). Korean Laws in English.

http://www.moleg.go.kr/english/korLawEng?.

Seoul Metropolitan Government. (2013).

http://urban.seoul.go.kr/DUPIS/sub3/sub3_1.jsp.

Seoul Metropolitan Government (2014). 2012 Seoul Survey http:

//stat.seoul.go.kr//pdf/e-webzine75/pdf.

Seoul Metropolitan Government. (2015). Seoul Statistics in 2012

http://stat.seoul.go.kr/.

Statistics Korea (2015). http://www.kostat.go.kr/eng/.

Supreme Court of Korea http://www.courtauction.go.kr/.

104 No, Kim and Yu

Appendix The Court Auction Process of Residential Real

Estate in Korea

In Korea, the court auctions of residential real estate are exclusively controlled

by the district courts in accordance with the process shown in Figure 1. Under

Korean laws, there are two types of auctions available to creditors: compulsory

judicial and court-administered auctions. The former is where the creditor

obtains a monetary judgment from a court against a debtor but the debtor does

not voluntarily address the judgment or order. The latter is where the creditor

holds a lien or mortgage on a relevant real estate property of a debtor and the

debtor defaults on his/her obligations. These two kinds of auction processes are

very similar.

The court auction process of residential real estate in Korea can be divided

into three stages: pre-bidding, actual bidding, and post-bidding. In the pre-

bidding stage, the courts order the appraisers to estimate the current price of the

auction items. The courts must then set the initial reserve prices based on the

reported value from the appraisers. After the filing period, the courts publicly

announce the auction date and reserve prices in detail for two weeks in

newspapers and on the webpage of the Supreme Court of Korea

(http://eng.scourt.go.kr). During the actual bidding stage, the courts receive

sealed bids from the auction participants with a 10% deposit of the reserve

prices. Bidders usually establish their bids by adding a certain amount of money

to the reserve price based on their own evaluations. Accordingly, the reserve

price in all practicality assumes the important role of a reference point. The

courts then close the bidding process after an hour. The highest bidder wins the

auction and must pay the bid amount in full within 30 days. If the bidders do

not meet the reserve prices, the courts continue the process at one-month

intervals, and reduce the reserve prices by 20–30% each time. During the court

auction process, bidders do not know the bids of other competitors. Therefore,

the court auction process of residential real estate in Korea is a first-price

sealed-bid type of auction. Finally, the post-bidding stage deals with the

payment of the purchase price and distribution of the proceeds after the

authorization of the sale. The deadline for the payment period will be generally

set as one month after the authorization of the sale (Attorneys at Law

YULCHON, 2012).